Dictation glossary

Filler-word removal

Filler-word removal means deleting spoken hedges — um, uh, like, you know — from a transcript so the remaining text reads more like writing than like a raw recording. Speech-to-text will happily spell those tokens if you said them. Cleanup then drops them, sometimes with simple word lists and sometimes with a broader rewrite. It is a post-processing step, not a change to how the microphone is sampled.

Get WhisperJot Try the browser demo
In more detail

What is Filler-word removal?

People fill time while they think. In conversation that is natural; in an email it looks unedited. Filler-word removal is the first, most mechanical cleanup pass: match known fillers, delete them, fix leftover spaces and punctuation. Aggressive lists can eat words you meant, such as a quoted 'like' in a technical sentence. Conservative lists miss 'sort of' and false starts. The right default is boring: drop the obvious vocal stumbles and leave the rest for you or for a later language-model polish.

Filler removal is one slice of text cleanup, alongside punctuation, capitalization, and custom replacements. It does not fix wrong names and it does not add structure you never spoke. Voice commands such as 'new paragraph' are the opposite idea: they insert form rather than deleting noise. If a product advertises 'ready to send' text, filler-word removal is usually doing a large share of that claim, because ums are frequent and easy to detect compared with true recognition errors.

For writers who speak

Why it matters for dictation

Dictation copies how you talk. Without filler removal, every message needs a cleanup pass that cancels the time you saved by speaking. With it, short dictation can go out as-is. The feature is small, visible, and easy to over-tune, so it is a good test of whether a vendor treats post-processing as a real part of the writing loop.

In this product

How WhisperJot handles it

WhisperJot removes filler words as part of automatic cleanup, then applies punctuation and optional local LLM polish on 127.0.0.1. That pass runs after recognition, whether you used Jot Local, Jot Local Pro, or the opt-in Jot Cloud engine. The local engines remain the default for transcription itself.

Related terms

Keep reading in the glossary.

Text cleanup

Text cleanup means editing a raw speech-to-text transcript so it looks like written language: punctuation, capitalization, filler-word removal, and user-defined replacements.

Read Text cleanup →

Custom vocabulary

Custom vocabulary means a user-supplied list of words, names, and spellings that a speech-to-text system should prefer when it hears something similar.

Read Custom vocabulary →

Voice dictation

Voice dictation is a way of writing in which you speak and software turns those words into editable text in an application.

Read Voice dictation →

Live partial transcription

Live partial transcription means showing in-progress text while audio is still being captured, instead of waiting until you stop talking to reveal a finished transcript.

Read Live partials →
Questions

Straight answers.

See also: Pricing, Use cases.

What is filler-word removal?

Filler-word removal is post-processing that deletes spoken hedges such as um, uh, and you know from a transcript. The recognizer may have spelled them correctly; cleanup decides they should not appear in the inserted text. It makes dictation look closer to typing. It will not repair a wrong content word, and a too-greedy list can delete a word you intended to keep.

Does filler removal change what the speech model heard?

No. The acoustic model still produces a token sequence. Filler removal edits that sequence afterward, like a copy editor. That is why you can sometimes see um in a live partial and then watch it disappear in the final insert. If you need a verbatim record, turn the pass off or use a transcription mode that preserves disfluencies on purpose.

Should I remove every like and basically?

Not always. Some fillers are content, especially in quoted speech or when you are discussing the word itself. Start with the obvious vocal stumbles and add phrases only if they show up in your sent mail. Pair the list with punctuation cleanup rather than trying to make one regex do a full rewrite. Optional language-model polish can catch what a static list misses, at the cost of more local compute.

One hotkey, any focused app.

Private local transcription by default, with an optional opt-in cloud engine.

Get WhisperJot — $99/year